Tech • AI • Robotics • Game

ENFR

Daily Podcast full article

Giga and Argentum commit $1.63B to the AI data-center race

Two fresh infrastructure commitments — Giga Energy’s $1 billion Texas lease and Argentum AI’s $630 million collaboration with Kyndryl — show how the AI buildout is moving from press-release ambition to contracted capacity, power-ready campuses and GPU-backed operating partnerships.

Generated November 18, 2026 at 12:19 PM≈ 7 min read
AI-generated illustration

Two deals, one message: AI capacity is becoming contractual

Giga Energy and Argentum AI have put a combined $1.63 billion behind new AI infrastructure commitments, giving the market another snapshot of how compute capacity is being assembled in the United States. Giga announced a 10-year, $1 billion lease for Project Clay, a 60-megawatt AI data center in Hunt County, Texas, with an undisclosed U.S.-based AI cloud provider . Argentum AI and Kyndryl announced a 20-year, $630 million data-center collaboration for an initial 10 megawatts of capacity at Kyndryl’s Boulder, Colorado facility, where Argentum is deploying Nvidia B300 systems for AI inference .

The two agreements are not identical. Giga’s deal is a long-term lease tied to a specific campus, a specific geography and a named power scale. Argentum’s is a collaboration that links a GPU cloud provider, an infrastructure operator and Nvidia systems inside an existing Kyndryl facility. But taken together, they point in the same direction: the AI market is moving from abstract demand forecasts toward contracted electrical capacity, rack-ready sites and operating partnerships.

Giga’s Texas lease turns demand into a site

Giga’s Project Clay is not merely a speculative data-center announcement. The company says it has already broken ground, will develop, own and operate the Hunt County campus for the customer, and expects the site to be ready for service in spring 2027 . That matters because a lease converts a broad AI-demand story into a more concrete operating plan. A cloud provider has committed to a facility; the developer has committed to a site; and the local grid, construction schedule and cooling design become part of the economics.

The scale is significant but also revealing. Project Clay is planned as a 60 MW GigaBase AI data center on roughly 30 acres, according to Giga . Texas Construction Hub reported that the $1 billion figure represents the value of the 10-year lease, not an announced construction contract, which is an important distinction when comparing this deal with traditional capex headlines . Spread evenly across the term and stated capacity, the lease implies about $16.7 million per megawatt over 10 years, or roughly $1.67 million per megawatt-year, though that simple arithmetic does not disclose the tenant’s total compute cost or Giga’s construction margin.

The bigger strategic point is that electricity access is becoming as decisive as chips. Giga says Project Clay is designed around dedicated electrical infrastructure that taps underutilized capacity and does not draw on distribution capacity serving existing homes and businesses . That framing directly addresses one of the central tensions around AI data centers: communities want tax revenue and jobs, but they do not want residential power bills, grid upgrades or water stress shifted onto them.

Water, speed and local politics enter the compute equation

Giga is also presenting Project Clay as a response to concerns about water use. The company says the site will use a fully closed-loop cooling system that operates independently of local municipal or groundwater supplies for cooling IT equipment, with local water use limited to sinks and employee bathrooms . Whether every claim survives long-term operational scrutiny will depend on execution, but the signal is clear: AI data-center developers now have to sell not just capacity, but community compatibility.

Construction speed is another selling point. Giga says the project will finish in under eight months, less than half the time of traditional builds, using a factory-first construction model in which much of the facility is manufactured off-site before final assembly . Texas Construction Hub added that Giga describes the project as a turnkey modular facility and reported that the company’s Hunt County project page says about 90% of construction labor happens in Giga factories rather than on site . In AI infrastructure, months matter: a facility delivered in spring 2027 can serve a different market than one stuck in permitting, interconnection or supply-chain queues until 2028.

The local economic package is also part of the pitch. Giga says Project Clay is expected to create about 300 construction jobs, 24 permanent full-time positions and millions in recurring annual tax revenue for local public needs . Those numbers are modest relative to the headline value of the lease, but they are central to how AI data centers are being negotiated with local communities.

Argentum and Kyndryl show the GPU-rack model

Argentum’s agreement with Kyndryl highlights the other side of the AI infrastructure stack: not just land and power, but GPU-ready operating capacity. Argentum and Kyndryl said the $630 million collaboration covers an initial 10 MW at Kyndryl’s Boulder, Colorado facility and supports deployment of Nvidia GPU systems . The companies said Argentum is deploying Nvidia B300 systems in a Tier III environment for demanding AI applications at scale, and that deployment is already underway .

That structure shows how AI capacity is increasingly built through layered partnerships. Argentum describes itself as a premium GPU cloud provider that originates, finances and operates GPU clusters, delivering dedicated bare-metal GPU capacity to model builders, inference providers and enterprises . Kyndryl contributes the mission-critical infrastructure and services platform, with the companies emphasizing power density, reliability and enterprise-grade operations . Nvidia is not the contracting counterparty in the announcement, but its B300 systems are the hardware around which the capacity is being organized .

The 20-year term is especially notable. AI hardware generations move quickly, but data centers, power arrangements and enterprise service relationships are long-lived. By locking in capacity at Kyndryl’s Boulder facility, Argentum is giving its customers a faster path to high-performance inference capacity while anchoring part of its business model in a long-duration infrastructure agreement . That is the new party guest in the enterprise stack: the GPU rack, accompanied by power contracts, cooling designs and managed-service obligations.

Why the combined $1.63 billion matters

The headline number is simple: $1 billion from Giga plus $630 million from Argentum equals $1.63 billion in newly announced AI-infrastructure commitments. But the importance is less about the arithmetic than the form of the commitments. One deal is a lease on a physical Texas facility. The other is a 20-year collaboration for GPU deployment inside a Kyndryl data center. Both attach AI demand to long-term operating assets.

That is a shift from the earlier phase of the AI boom, when many announcements were measured by theoretical model potential, future chip orders or broad “strategic partnership” language. These deals are more grounded. They name megawatts, facilities, operating roles, cooling systems and service windows. They also expose the bottlenecks now shaping the market: not only access to accelerators, but also power density, grid interconnection, site readiness, local approvals and water management.

There are caveats. Giga has not named the cloud provider behind the Project Clay lease . Argentum and Kyndryl have not named the fast-growing enterprise AI inference company described as the flagship anchor tenant at the Boulder facility . The public materials do not fully break down construction cost, financing terms, GPU procurement economics or customer pricing. So these announcements should not be read as complete project financial models.

Still, they are useful indicators. Giga’s lease suggests that AI cloud demand is strong enough to support long-term commitments to purpose-built power and real estate in Texas. Argentum’s collaboration suggests that GPU cloud providers are willing to secure enterprise-grade facilities for years, not quarters. For Nvidia, each deployment of named GPU systems inside long-lived facilities reinforces how accelerator choice can shape infrastructure design, power density and service relationships far beyond the initial hardware purchase.

The new buildout logic

The American AI data-center buildout is becoming a coordination exercise among three groups: workload owners that need compute, infrastructure operators that can deliver reliable facilities, and hardware suppliers whose systems define rack density and thermal requirements. Giga’s Project Clay sits closest to the land-power-construction layer. Argentum’s Kyndryl collaboration sits closer to the GPU-capacity and enterprise-service layer. Together, they show that AI infrastructure is no longer just a chip story or a cloud story; it is a power, property, cooling and operations story.

That is why these two commitments belong in the same frame. Giga’s Texas lease shows demand being anchored to a readying campus. Argentum’s Kyndryl deal shows demand being anchored to Nvidia-equipped capacity in a Tier III facility. The market is not simply buying more GPUs. It is installing the buildings, megawatts and operating contracts that let those GPUs become a service.

Comments

Be the first to comment.

Sources from the last 72 hours

  1. [1]Argentum AI signs $630M data center collaboration with Kyndryl to deploy NVIDIA GPUsNov 17, 2026, 1:00 AM
  2. [2]Giga Energy Inks 10-Year, $1B Lease on Project Clay, a 60MW AI Data Center in Hunt County, TXNov 17, 2026, 1:00 AM
  3. [3]Giga Energy Signs $1 Billion Lease for 60 MW Data Center in Hunt CountyNov 17, 2026, 1:00 AM
  4. [4]Argentum AI Signs $630M Data Center Collaboration With Kyndryl to Deploy NVIDIA GPUsNov 17, 2026, 6:00 PM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.